The new Tsunami 2's have a steam decoder that you can select a 3 cylinder chuff sound- triggered by BEMF, no cam or magnets.
According to a book called, "New Haven Railroad: A Fond Look Back", the 3-cylinder engines drove the NH shop forces crazy until they learned they were supposed to sound out of balance.
The issue is that the center piston, due to have a shorter and straight exahust path out the stack, sounded louder than the two outside pistons.
It was described in the book by an old timer who worked on the 3-cylinder engines as sounding like, "A washing machine, a wishing machine, a washing machine, a wishing machine" and so on.
It would take an interesting decoder to replicate that.
Overmod, given that the single-pickup flywheel supplied by the factory provides 22 pulses per driver revolution, no fix to modify the flywheel would help, only to the gear ratio.
24:1 gears (which most BLI's have) would allow the Paragon to work properly if its countdown firmware were reset from six to four. Or 18:1 gears and countdown of three, sacrficing some low-speed smoothness. But 22:1 would require more complex circuitry.
Your mechanical method sounds good. Epoxy is simple tho: I can set the exact phase of the chuff by experimenting with one magnet and the reed switch position, then attach the others to match either an even tempo, or slightly syncopated (mine). The magnets and switch can't normally be seen.
I used the same technique on a BLI "stealth" Dreyfuss Hudson. BLI was advertising that the "stealths" were upgradable (flywheel magnet already installed and place for speaker), but then they wouldn't sell me the Paragon upgrade kit. Annoying.
Hal
I keep trying to add something but the site keeps crashing the page and losing the content.
This is a really good approach. I suggest enhancing it slightly by making a flat ring out of plastic, the thickness of the magnets, transparent if visibility is important , and drilling indexed holes for the magnets; that will provide a smooth surface for a 'contact' sensor (allowing a Hall type with elastomer 'spring' to stay in near-contact without binding or wear; permit easy attachment and re-indexing for quartering; allow offset and even circumferential adjustment for the 'magnitude' of inside-cylinder Gresley-gear wear or poor adjustment.
I think it may be possible to have multiple 'tracks' as in hard drives so the inside-cylinder magnets could be read separate from the outside-cylinder ones if there are differences in sound or release time.
Occurs to me you could also provide a circumferential ring on the flywheel, machined to precise OD for an exact count, with whatever number of magnets gives a modulo-6 count per physical driver revolution at the decoder. Machine a groove in the flywheel if desired. Here, parallel circumferential tracks like those on an old drum memory setup would give the differential between cylinders, or multiple individual signals for articulateds including actual compound Mallets with accurate sound for bypass-valve actuation, etc.
If you want precise driver-driven chuffs, I found a way that works for me.
I have a BL UP 4-12-2. The problem was that the Paragon2/3 cannot syncronize six chuffs per revolution with the gearbox in that engine because the pickup on the flywheel must be counted down, and IIRC, the gear ratio is 22:1. Apparently, BL didn't realize this, and had to ship them wrong; they admitted there was no fix. So I replaced the P3 with a WOW decoder and epoxied six tiny disk magnets from KJ Magnetics on the inside of the front left driver. I used a subminiature reed switch for the pickup, mounted on the frame. I tried a Hall effect sensor first, but its spacing was too critical for a driver that has to have some sideways freedom. I like this approach because, unlike a cam, there is nothing to wear or need cleaning.
Varying the spacing of the disks can get you some syncopation if you want. On-line recordings of the 4-12-2's reveal mis-timing, probably because the crews hated adjusting the middle cylinder from a pit, and tried to get away with doing it as seldom as possible.
You can tune the chuff rate with most decoders today. Instead of tuning it for four chuffs per revolution, why not just adjust it for 6 chuffs per revolution ?
.... Or am I missing something in how the proto-type functioned ?
Mark.
¡ uʍop ǝpısdn sı ǝɹnʇɐuƃıs ʎɯ 'dlǝɥ
Forum member Riley B asks:
I would like to install DCC sound in my SP 4-10-2. The problem is that it has three cylinders. Would a custom made disk mounted to a driver be the way to go (how would I do this ?) or is there a simpler solution? I would really appreciate some guidance on this one. Thank you for your help.
He posted this in General, but got locked and redirected here by moderation. Evidently he didn't get the memo that this was done, and will come back to find he has to ask all over again. This is to get a head start on actual answers to his question here.
I'd start by observing that at least some sound decoders synthesize the acceleration of 'chuffs' by the voltage applied, and not with a mechanical cam or disk actuating a reed switch (as was ancient practice, involving a stove-bolt square nut - or hex nut for three cylinders - as a cheap cam in the good old days). As a 'default' solution, since the SP 4-10-2s were simple engines with Gresley conjugation we could start by assuming all six exhaust events per revolution are equal, and find a way to speed up the voltage-dependent rate by 1/3. Alternatively someone might edit the sound file for a three-cylinder locomotive (perhaps a British prototype) with the correct exhaust sounds or hooked-up varieties to work correctly, and patch that into a dataset otherwise correct for the American prototype.
He could indeed use an optical encoder as a 'trigger wheel' approach to get the timing off a driver axle directly, and this would give accurately coordinated 'timing' events at any rotational speed if done right. The initial issue I see is how to get the encoder disk properly registered on the axle without removing a driver and having to requarter afterward. He'd need to split and install the wheel, time it appropriately, and then attach it reasonably firmly to the axle, then find a proper place for the optical pickup that reads transversely through the slots in the wheel. Note that concentricity is much less a problem for comparatively long radial slots, but circumferential spacing of the 'leading edge' of the slots will be important.